The practice of holding contests and awarding prizes to develop specific innovations has been a proven method for centuries. In 1795, the French government offered 12,000 francs to whoever could come up with a food preservation method that would allow Napoleon to feed his army abroad. After 15 years of failed experiments, Nicolas Appert, a Parisian chef and confectioner, solved the problem by boiling food after it had been sealed in airtight containers, and became known as the father of canning.
In recent decades, the use of innovation contests has surged. In 2006, Netflix offered $1 million for an algorithm that improved its movie recommendation system’s performance by 10 percent; the prize was claimed in 2009 by a multinational collective known as BellKor’s Pragmatic Chaos. Still open? The Google Lunar XPRIZE, which will pay a $20 million grand prize to the first team that lands a privately funded robot on the moon and sends mooncasts back to Earth.
Designers of innovation contests have two main elements to consider: how to allocate prize money and whether to regularly disclose information about contestants’ successes. There are two salient options for prize allocation: the winner-takes-all method, which awards the entire prize to the first contestant who succeeds, and equal sharing, which divides the prize between all contestants who come up with a solution by a deadline. Likewise, there are two main options for information disclosure: a public information structure, in which any contestant’s success is announced immediately, and a hidden information structure, in which no one knows whether anyone else has succeeded until the deadline.
But which of these choices is more likely to motivate contestants? Professor Marina Halac, working with Navin Kartik and Qingmin Liu, professors of economics at Columbia University, analyzed these and other options in a recent study. Most people assume that a winner takes all/public information contest is the optimal design, Halac says. “There’s truth to that, because if you’re investing your time and effort to develop an innovation, you want to know that you’ll win the full prize if you’re successful,” she explains. “And you also want to know if anyone else has already come up with a solution.”
Through the use of mathematical models, Halac and her coauthors found that a winner takes all/public information contest is ideal under some, but not all, conditions. The exception: when there is uncertainty about the possibility of success. “Sometimes the result is impossible to obtain,” Halac says. “Suppose you’re trying to come up with a new pharmaceutical, for example. It might not be feasible given the current technology or within the relevant time frame.” Uncertainty about the possibility of success leads to pessimism over time: the longer contestants work on a problem but do not come up with any results, the more likely they are to quit. “If you’re competing in a public contest and know that no one has come up with a solution, you’ll become very pessimistic about its viability,” she says. “Eventually, you’ll decide that it’s too costly to keep working.”
To keep motivation high when the probability of success is low, it is best to use a hidden information structure — specifically, not disclose contestants’ lack of success. But hiding information cannot help if the contest is winner takes all, as in that case, any competitor who succeeded would have also taken the entire prize. The ideal design when success is unlikely is an equal sharing/hidden information contest, Halac says. “Even though the equal sharing reduces incentives — because anyone who succeeds might have to share the prize — this design prevents contestants from becoming pessimistic too soon.” More generally, an optimal design combines the options: opening the contest with a winner takes all/public information structure, and then switching to equal sharing/hidden information at a predetermined point. “The more unlikely it is that anyone will succeed, the earlier the contest should switch to an equal sharing/hidden information design,” explains Halac. “When the chances of winning are very low, less information is a good thing.”
About the researcher
Professor Halac's research studies dynamic incentive provision in settings with private information, contracting constraints, and learning. Halac teaches an elective course on Game...Read more.